Enregistré dans:
Détails bibliographiques
Auteurs principaux: Fei, Zhengcong, Li, Debang, Qiu, Di, Wang, Jiahua, Dou, Yikun, Wang, Rui, Xu, Jingtao, Fan, Mingyuan, Chen, Guibin, Li, Yang, Zhou, Yahui
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2504.02436
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866909563371913216
author Fei, Zhengcong
Li, Debang
Qiu, Di
Wang, Jiahua
Dou, Yikun
Wang, Rui
Xu, Jingtao
Fan, Mingyuan
Chen, Guibin
Li, Yang
Zhou, Yahui
author_facet Fei, Zhengcong
Li, Debang
Qiu, Di
Wang, Jiahua
Dou, Yikun
Wang, Rui
Xu, Jingtao
Fan, Mingyuan
Chen, Guibin
Li, Yang
Zhou, Yahui
contents This paper presents SkyReels-A2, a controllable video generation framework capable of assembling arbitrary visual elements (e.g., characters, objects, backgrounds) into synthesized videos based on textual prompts while maintaining strict consistency with reference images for each element. We term this task elements-to-video (E2V), whose primary challenges lie in preserving the fidelity of each reference element, ensuring coherent composition of the scene, and achieving natural outputs. To address these, we first design a comprehensive data pipeline to construct prompt-reference-video triplets for model training. Next, we propose a novel image-text joint embedding model to inject multi-element representations into the generative process, balancing element-specific consistency with global coherence and text alignment. We also optimize the inference pipeline for both speed and output stability. Moreover, we introduce a carefully curated benchmark for systematic evaluation, i.e, A2 Bench. Experiments demonstrate that our framework can generate diverse, high-quality videos with precise element control. SkyReels-A2 is the first open-source commercial grade model for the generation of E2V, performing favorably against advanced closed-source commercial models. We anticipate SkyReels-A2 will advance creative applications such as drama and virtual e-commerce, pushing the boundaries of controllable video generation.
format Preprint
id arxiv_https___arxiv_org_abs_2504_02436
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SkyReels-A2: Compose Anything in Video Diffusion Transformers
Fei, Zhengcong
Li, Debang
Qiu, Di
Wang, Jiahua
Dou, Yikun
Wang, Rui
Xu, Jingtao
Fan, Mingyuan
Chen, Guibin
Li, Yang
Zhou, Yahui
Computer Vision and Pattern Recognition
This paper presents SkyReels-A2, a controllable video generation framework capable of assembling arbitrary visual elements (e.g., characters, objects, backgrounds) into synthesized videos based on textual prompts while maintaining strict consistency with reference images for each element. We term this task elements-to-video (E2V), whose primary challenges lie in preserving the fidelity of each reference element, ensuring coherent composition of the scene, and achieving natural outputs. To address these, we first design a comprehensive data pipeline to construct prompt-reference-video triplets for model training. Next, we propose a novel image-text joint embedding model to inject multi-element representations into the generative process, balancing element-specific consistency with global coherence and text alignment. We also optimize the inference pipeline for both speed and output stability. Moreover, we introduce a carefully curated benchmark for systematic evaluation, i.e, A2 Bench. Experiments demonstrate that our framework can generate diverse, high-quality videos with precise element control. SkyReels-A2 is the first open-source commercial grade model for the generation of E2V, performing favorably against advanced closed-source commercial models. We anticipate SkyReels-A2 will advance creative applications such as drama and virtual e-commerce, pushing the boundaries of controllable video generation.
title SkyReels-A2: Compose Anything in Video Diffusion Transformers
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2504.02436